Formatter can insert mysterious extraneous parentheses in with statements

[lpulley]

My ruff --version is 0.7.1.

I've encountered a case where ruff format adds nesting parentheses for seemingly no reason:

with (
    open(
        "some/path.txt",
        "rb",
    )
    if True
    else open("other/path.txt")
    # Bar
):
    pass

➜ ruff format --diff up034.py
--- up034.py
+++ up034.py
@@ -1,10 +1,12 @@
 with (
-    open(
-        "some/path.txt",
-        "rb",
+    (
+        open(
+            "some/path.txt",
+            "rb",
+        )
+        if True
+        else open("other/path.txt")
     )
-    if True
-    else open("other/path.txt")
     # Bar
 ):
     pass

1 file would be reformatted

i.e. it wants this:

with (
    (
        open(
            "some/path.txt",
            "rb",
        )
        if True
        else open("other/path.txt")
    )
    # Bar
):
    pass

UP034 complains about this, for good reason.

If the # Bar comment is removed, the formatter makes no changes. More curiously, if I keep # Bar but add as _, the formatter makes no changes. (I also checked that Black does not do this.)

I've resolved the issue in the meantime by moving the # Bar comment to a different line, but this seems like a formatter bug.

[MichaReiser]

Thanks for reporting this. Agree, this looks silly but I'm a bit scared from fixing it... with item formatting is complicated.

[lpulley]

Hmm... with Ruff 0.7.2, the example I gave above is still problematic; it still wants to add extra ( ) and fail UP034.

[MichaReiser]

Hmm... with Ruff 0.7.2, the example I gave above is still problematic; it still wants to add extra ( ) and fail UP034.

Yeah, we can't fix it before promoting the new style to stable. You would have to use format.preview = true for now

[lpulley]

Aha; I missed that. Thanks for the clarification!!

…

-------- Original Message --------

On 11/4/24 13:52, Micha Reiser wrote: > Hmm... with Ruff 0.7.2, the example I gave above is still problematic; it still wants to add extra ( ) and fail UP034. Yeah, we can't fix it before promoting the new style to stable. You would have to use format.preview = true for now — Reply to this email directly, [view it on GitHub](#14001 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/ABW4EY3TB6AAI6UTPRAK5FDZ67GBFAVCNFSM6AAAAABQ4NIQLOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJVGU3TKNZTHE). You are receiving this because you authored the thread.Message ID: ***@***.***>

[MichaReiser]

You can subscribe to #13371 to get notified when the preview style gets stabilized.

[lpulley]

I'm trying out 0.9.0 and still seeing this behavior. Did this make it into the 2025 style?

[MichaReiser]

Hmmm, no. That's on me.

[lpulley]

Not the end of the world; just wanted to make sure I wasn't missing anything obvious. Thanks.

[MichaReiser]

Thanks for being so understanding and sorry that I missed that one last preview style

[cj81499]

Hey @MichaReiser, I'm one of @lpulley's coworkers. Will this make it into 0.9.1? Or will we be waiting till the 2026 style?

fwiw, we've been super happy with ruff and this will not change that :) Thanks for helping to build such an awesome piece of software!

[MichaReiser]

Thanks for the kind words.

We might want to pull it into 0.10, considering that changes should be super rare. I don't feel comfortable shipping it as part of a patch release because it would violate our versioning policy

[cj81499]

Sounds good to me, thank you!